Inference Engine

The inference engine uses the structure within the knowledge graph to identify data patterns and to infer insights using a symbolic approach.

Symbolic Structures
Babelfish converts business data to a string, using an explicit symbolic template (refer Slide 9 in the below deck), which is derived based on the relationships within the graph. The string can be explained as alphanumeric representations, which are short codes that represent the relationships of a given entity

State Activation
The strings are dynamic in nature. Every time new data is captured, it computes for symbol specific weights which are used to activate or change associated states. An illustration of the various streaming processes that update the string can be viewed on Slide10

Data Selection
When a user asks for a query, based on key semantic roles, matching strings are extracted. Based on the context of the query, which could be descriptive, predictive or prescriptive in nature, the strings are subjected to inferencing rules to output data based on the context. Check Slide 13/14


The below deck illustrates how the backend can be automated to get the contextual data search up and running.